Abstract:
© 2017 IEEE. The automatic extraction of drug side effects from social media has gained popularity in pharmacovigilance. Information extraction methods tailored to medical subjects are essential for the task of drug repurposing and finding drug reactions. In this article, we focus on extracting information about side effects and symptoms in users' reviews about medications in Russian. We manually develop a real-world dataset by crawling user reviews from a health-related website and annotate a set of reviews on a sentence level. The paper addresses the classification problem with more than two classes, comparing a simple bag-of-words baseline and a feature-rich machine learning approach.